ABSTRACT
A quick glance through some recent volumes of the communication literature reveals that communication researchers often advance hypotheses focused on such differences between people, conditions, or stimuli. The most commonly employed method of comparing group means requires the “pooling” of the sample variances of each group in order to estimate the standard error of the mean difference. The pooled variance approach to testing the equality of means is particularly robust to violations of homogeneity of variance when the sample sizes are unequal. Causal inferences of this sort can be made in well designed-experiments even in the absence of random sampling. Whereas random assignment affords causal inferences, random sampling affords population inferences. The most common solution to this apparent dilemma is to just ignore the fact that the sample t test was designed for an entirely different problem and pretend as if the sample is a random one.
